This is a resubmission application for a K01 award for Tara Lagu, MD, MPH, a board-certified internist and Assistant Professor at the Tufts University School of Medicine. Dr. Lagu is based in the Center for Quality of Care Research (CQCR) at Baystate Medical Center (BMC) in Springfield, Massachusetts. Dr. Lagu's focus is on the quality and outcomes of care for patients with heart failure (HF). Her long-term goal is to become an independent cardiovascular outcomes researcher. Her immediate goals are 1) to use quantitative and qualitative methods to define optimal clinical strategies for the care of critically ill patients with HF and 2) to develop, through mentorship and formal coursework, skills that will help her to become an independent investigator. Her training as a Robert Wood Johnson Clinical Scholar and experience as a young faculty member at BMC will help her to accomplish these goals. Dr. Lagu has assembled a mentorship team with expertise in both the proposed work and in helping junior faculty transition to independence. Her primary mentor is Dr. Peter Lindenauer. Her co-mentor is Dr. Harlan Krumholz. Consultants include Dr. Penelope Pekow (biostatistician) and Dr. Kathleen Mazor (psychometrician). Dr. Lagu will have full access to the resources of both BMC and Tufts University School of Medicine. The CQCR (directed by Dr. Lindenauer, in collaboration with a group of senior faculty), which will provide mentorship, biostatistical, and administrative support as well as a peer group of junior faculty with similar interests. This research proposal aims to answer the question: What are the clinical strategies that enable hospitals to achieve lower risk-adjusted mortality rates for critically ill patients wih HF? Critically ill HF patients have a 30-day mortality rate as high as 20%, indicating that these patients contribute disproportionately to reported hospital variation in HF risk adjusted mortality rates, yet there are no large observational studies of these patients. We will first develop a mortality prediction method for use in observational studies of critically ill HF patients that taks advantage of a unique and powerful dataset. In response to reviewer comments, we have included markers of HF etiology in this model and will separately validate the model in subgroups with known HF etiology. Next, we will examine variation in risk-adjusted mortality and will determine the proportion of variation that can be explained by differences in use of evidence-based clinical practices. Because we anticipate that most of the observed variation in risk-adjusted outcomes will remain unexplained, in Aim 3 we will develop qualitative methods that can be used to identify the clinical strategies and practices employed by hospitals with low risk-adjusted mortality rates for this population. These methods will generate preliminary data for an R01 that will include qualitative interviews from many more hospitals and will validate the hypotheses through a random sample questionnaire linked to hospital performance.